GitHub connectivity issues | Hacker News Reader

GitHub connectivity issues

status.github.com

43 pointselectic19 comments

wow I knew it. Everything is absolutely crawling. I am trying to push up my repo and get some packages. I could tell something was definitely up. Damn it. Thanks for the update.

Couldn't push, so came here to find out it wasn't only me. 'Server aborted the SSL handshake'

Edit: seems back to normal now

What I actually find more interesting is the spike in hook delivery time that correlated with the spike in exception rate.

My guess would be that some hooks were timing out which possibly causes an internal exception? Maybe? It is interesting that they would include timeouts of external services in their exception count. Purely speculation though.

Ruby with resque/sidekiq will definitely cause an exception if an http call fails or times out. This is what causes it to end up back in the job queue.

00:38 Eastern Standard Time All systems reporting at 100%.

https://status.github.com/messages

I'm working on GMT+3, and our work-day starts at 7am. Sort of sucks to have my core work hours be the typical SFO downtime slot...

Looks like I wasn't the only one! Wonder who's behind this one.

It's completely unusable for me :(

works for me now.

How do we know it's a DDOS and not some sort of operational failure? Mods should change the title until it is confirmed.

Agreed. Exact text on status page right now:

> 20:51 PDT - We are investigating reports of connectivity issues to GitHub.com.

No mention of DoS or DDoS.

Ok, we changed the title from "GitHub under DDOS."

Look at the traceroute. This is a DDoS mitigated via Prolexic.

Wouldn't the traffic run through to Prolexic regardless if it was under a DDOS or not?

Typically, DDoS mitigation services offer a limited number of 48-72 hour mitigations per year (can be a large number, depending on the contract). Outside of these periods the traffic is not routed via them.

How does that work? Does the DDOS mitigation software change DNS and if so how can it change the DNS in the middle of an attack assuming the attackers?

Only very basic DNS mitigations services have you change DNS and proxy through them. Predictably GitHub is not using one of those.

If you really want to protect your service (your own DNS or say SMTP, WebSockets, TCP), then you need to change the internet routing. This is done via BGP announcements of your IP subnets, such that you announce your inbound routes via the mitigation providers. The providers scrub the traffic and deliver clean traffic via a GRE tunnel to your routers. The outbound traffic is routed directly via your upstream providers.

Reply on news.ycombinator.com